An Affective Engineering Study of Vibrational Cues and Affect
Click here to load reader
Transcript of An Affective Engineering Study of Vibrational Cues and Affect
7/27/2019 An Affective Engineering Study of Vibrational Cues and Affect
http://slidepdf.com/reader/full/an-affective-engineering-study-of-vibrational-cues-and-affect 1/8
An Affective Engineering Study of Vibrational Cues and Affect
When Touching Car Interiors
Louise Manfredi†, Brian Henson, Cathy Barnes, Tom Childs
Affective Engineering Laboratory
School of Mechanical Engineering
University of Leeds
Leeds LS2 9JT
United Kingdom
Keywords: affective engineering, kansei engineering, touch.
Category: Research paper
INTRODUCTION
Within the field of psychology, and psychophysics in particular, the understanding of texture and the perception of it has been an exciting research field since Katz’s The
World of Touch in 1925 (Kreuger, 1989). The World of Touch underpins the view
that movement and the resulting vibrations are almost certainly needed in texture
discrimination. Bensmaia and Hollins (2003) concluded in a recent study concerning
roughness and the speed of lateral finger movement that there was a relationship
between frequency and spatial periods on the surfaces tested within the context of
roughness. In contrast, Lederman (1982) found that vibration has no service in the
perception of roughness which leaves no real conclusive evidence as to the
importance of vibration as a whole. Although much work has been undertaken in the
perception of texture, little has been undertaken in the application of textures in
product design such as this study presents.
An affective engineering study is reported in this paper in which relationships
between tactile vibrational cues and affect were examined in the context of car
interiors. The motivation behind the research was to generate hypotheses to test in
future experiments concerning how touching surface textures evokes particular
emotional responses. A variety of textures for mould decoration on otherwise
identical black plastic tiles were chosen as stimuli. Any variation in the affective
response of participants is therefore due to the surface textures of the tiles. The
context was car interiors and the demographic was males aged 18 to 26. Focus groups
7/27/2019 An Affective Engineering Study of Vibrational Cues and Affect
http://slidepdf.com/reader/full/an-affective-engineering-study-of-vibrational-cues-and-affect 2/8
were used to generate and reduce adjectives using the triad method and stimuli were
chosen using semantic mapping. The semantic differential method was used to
generate a semantic space for the tiles, which was compared with the frequency
components of the vibrations when touched. The profiles of the frequency spectra
might be responsible for the emotional responses of participants when touching
surface textures.
METHODOLOGY
Two focus groups were held to generate adjectives. All participants were unaware of
the precise purpose of the focus groups, other than that they would be asked to talk
about their attitudes and feelings about the interiors of cars, and that it would be audio
recorded. Thirteen male undergraduate students took part in the focus groups.
The first activity in each session was used as an icebreaking exercise to encourage
conversation and interaction within the group who were not familiar with each other.
They were asked to talk about cars that they liked and disliked.
In the first focus group, adjectives were generated by using the triad method on pictures of cars (Thomson and McEwan, 1988). Six images of car exteriors and 6
images of interiors were mounted on A3 boards and randomly ordered. The images
of the car exteriors were shown to the group, three images at a time. The participants
were asked the similarities and differences between the three images and their
answers were audio recorded. The group was then shown the images of car interiors
in the same way. The two sets of images were arranged in a random order again, but
this time with both interiors and exteriors within each triad. Once again, the pictures
were shown to the group in triads and they were asked the similarities and differences.
Finally, the participants were shown the all 12 images again, one at a time, and asked
to choose one adjective that best described their feelings.
The triad method was used for the second focus group study and the stimuli used were
18 parts of car interiors, such as steering wheels, which were prearranged into triads
comprising of all aspects of the interior. Participants were encouraged to feel each
stimulus and, again, find similarities and differences between them.
After the focus group studies were completed, adjectives were collected from the
session recordings. Using adjective reduction guidelines (Barnes et al, 2007),
unsuitable words were rejected and the most popular and strongest words were
retained. Emotive words were grouped by similar meaning words using an affinity
diagram. Twelve words were selected: quality, masculine, engineered, warm, cheap,
exciting, comfortable, sophisticated, relaxed, unique, sporty and fashionable. The
opposite polarity of each word was indicated using ‘not’.
The stimuli for this study were samples of mould decorations purchased from Standex
(Standex). As supplied by Standex, the samples consisted of sheets of black plastic
moulded with a number of different surface textures (Figure 1). For this experiment,
the sheets were sawn into tiles, each with a different surface texture. From the 50
available tiles, 28 were selected by the authors. This was carried out in a subjective
manner with touch being the sole discriminator. Many of the tiles provided had the
7/27/2019 An Affective Engineering Study of Vibrational Cues and Affect
http://slidepdf.com/reader/full/an-affective-engineering-study-of-vibrational-cues-and-affect 3/8
same ‘pattern’ design but different grades of roughness, so it was decided that that for
this study, all pattern types with a variety of roughnesses be used.
The tiles were placed inside a black box, which had a front entrance with a small
window through which the participants could feel a section of each surface. The
boxes were black to match the colour of the tiles; none of which could be seen. This
ensured that the results yielded were based wholly on the tactual qualities of the tiles.
To reduce the number of stimuli, three semantic mapping exercises were carried out
against axes reading cheap-expensive, relaxing-exciting, each with two participants
who had not taken part in the focus groups. The participants who were asked in pairs
to determine where on the map they would best place the tiles, numbered 1 to 28 with
regards exclusively to the way the tile felt when touched. Each tile had a
correspondingly numbered marker to be placed on the map. The participants were
permitted to revisit their initial decisions and revise them if required and tile markers
could be placed on top of each other. The students were asked during the study to
keep the theme of car interiors in the forefront of their minds throughout the decision
making process. A photograph was taken of each of the completed maps and they
were compared. Tiles that were similarly placed on each of the three maps and arange of tiles from each of the maps’ quadrants were selected. This resulted in eleven
tiles (Table I).
Semantic differential questionnaires were prepared using the adjectives selected from
the focus group sessions. The words were presented in a random order and with
random polarity. Fifty seven male participants aged between 18 and 16 were
recruited in small groups to rate their subjective response to the eleven tiles on the
questionnaire. This activity took place in a controlled environment in a special
purpose affective evaluation room. Each group was read the same introduction sheet,
explaining the purpose of the study and what they were required to do. Incentives
were given as a token of appreciation. They were asked to fill out a practise sheet
with tiles not part of the experiment in order to familiarise themselves which the style
of questionnaire. The participants were asked to complete the questionnaire in silence
so not to disturb or influence the opinions of other participants. The data from the
semantic differential questionnaires was transcribed twice to spreadsheets, compared
and corrected. Missing responses were replaced with average values. The data was
analysed using principal components analysis in SPSS 14. Varimax rotation was used
and components with an eigenvalue greater than 1 were retained. The positions of the
tiles in semantic space were calculated.
7/27/2019 An Affective Engineering Study of Vibrational Cues and Affect
http://slidepdf.com/reader/full/an-affective-engineering-study-of-vibrational-cues-and-affect 4/8
Figure 1. Standex mould decorations.
Table I. Stimuli used in semantic
differential study
Tile Standex no.
1 MT 9022
2 MT 9015
3 MT 90974 MT 9081
5 K9000G
6 MT 9124
7 MT 9002
8 MT 9112
9 MT 9049
10 MT 9080
11 MT 9030
The tiles were mounted on a Kistler piezo-electric force measurement platform that
could respond to frequencies in sliding friction force up to at least 1kHz. Vibrationmeasurements of the 11 tiles were obtained by the principal author pulling the left-
hand index finger in a continual motion across each tile from right to left at a speed of
between 5 and 10 mms-1
. The data was captured from the Kistler measurement
apparatus on LabView recording 1000 times per second. Each data set lasted for
approximately 7 seconds. MatLab was used to apply a fast fourier transform to the
vibration data to give a graphical representation of the frequency spectra of the tactile
vibrations. The autocorrelation plots and the frequency spectra were then compared
with the positions of the tiles in semantic space.
RESULTS
At first, the principal components analysis was unable to generate a positive definite
correlation matrix. This was resolved by removing the averages for the word
‘fashionable’ from the analysis, because it correlated highly with the word
‘sophisticated’. Subsequent analysis identified three components (Table II).
Component 1 accounted for 49.2% of the variance; component 2, 25.8%; and
component 3, 14.0%. The words that loading most highly on component 1 were
‘exciting’ and ‘unique’; on component 2, ‘quality’ and ‘sophisticated’; and on
component 3, ‘warm’. The semantic space is illustrated in Figures 2 to 4.
The frequency spectra of the vibrations that occur when a finger is moved across the
surface of each texture are shown in Figures 5 to 15. The ordinate axis on each graph
is relative amplitude.
7/27/2019 An Affective Engineering Study of Vibrational Cues and Affect
http://slidepdf.com/reader/full/an-affective-engineering-study-of-vibrational-cues-and-affect 5/8
Table II. Rotated component matrixComponent
1 2 3
Quality -0.165 0.929 -0.094
Masculine 0.476 -0.491 0.691
Engineered 0.743 -0.153 -0.104
Warm -0.207 0.001 0.952
Cheap -0.210 -0.787 -0.095
Exciting 0.870 0.332 0.293
Comfortable -0.828 0.524 0.079
Sophisticated -0.111 0.901 -0.180
Relaxed -0.811 0.557 -0.033
Unique 0.947 0.145 -0.025
Sporty -0.544 0.675 -0.420
-2
-1
0
1
2
3
4
-2 0 2 4 6
5
7
910
1
8
11
3
4
6
2
PC 1
PC 2
Figure 2. Stimulus loadings incomponents 1 and 2.
-2
-1.5
-1
-0.5
0
0.5
1
1.5
-2 -1 0 1 2 3 4
6
2
4
3
7
9 118
10
1
5
PC 2
PC 3
Figure 3. Stimulus loadings in
components 2 and 3.
-2
-1.5
-1
-0.5
0
0.5
1
1.5
-2 -1 0 1 2 3 4
811
6
2
9
10
4
1 3
5
7
PC 1
PC 3
Figure 4. Stimulus loadings incomponents 1 and 3.
Figure 5. Spectrum of tile 1.Figure 6. Spectrum of tile 2.
7/27/2019 An Affective Engineering Study of Vibrational Cues and Affect
http://slidepdf.com/reader/full/an-affective-engineering-study-of-vibrational-cues-and-affect 6/8
Figure 7. Spectrum of tile 3.
Figure 9. Spectrum of tile 5.
Figure 11. Spectrum of tile 7.
Figure 13. Spectrum of tile 9.
Figure 15. Spectrum of tile 11.
Figure 8. Spectrum of tile 4.
Figure 10. Spectrum of tile 6.
Figure 12. Spectrum of tile 8.
Figure 14. Spectrum of tile 10.
7/27/2019 An Affective Engineering Study of Vibrational Cues and Affect
http://slidepdf.com/reader/full/an-affective-engineering-study-of-vibrational-cues-and-affect 7/8
DISCUSSION
The plots of the frequency spectra (Figures 5 to 15) for each surface differ from each
other in a number of ways. They can be characterised by:
• The number and distinctiveness of frequency peaks.
• The extent to which frequency peaks exist above 1.8Hz.
• The steepness and radii of the curves that results if a line is drawn connectingeach frequency peak.
• The frequency of the maximum peak.
• The range and relative amplitude of low frequency (>0.5Hz) components.
Consideration of the relationships between the positions of the stimuli in the semantic
map and their frequency spectra is notable for how surfaces with similar spectra elicit
different responses, and for how surfaces with different spectra can elicit similar
responses.
Against component 1, which is characterised by feelings of uniqueness and
excitement, surfaces 2 and 11 appear at opposite ends of the response scale, but have
very similar spectra. On the other hand, surface 6 is close to surface 2 in component
1, but they have very different spectra. Surfaces 2 and 6 are only similar in the range
of low frequency components, but surface 11 is also similar in this respect. Surfaces
1, 4, 5 and 11 evoke a low response against component 1, and are very similar in the
distinctiveness of their frequency peaks and the presence of frequency peaks above
1.8Hz, but if this might indicate a relationship then the position of surface 2, which
also shares these attributes, is anomalous.
Against component 2, surfaces 5 and 2 appear at opposite ends of the response scale,
but have very similar spectra. These two surfaces differ in that surface 2 has a more
pronounced frequency peak close to 0.7Hz, and so does surface 4, which is close to
surface 2 in semantic space. It is difficult to conclude a relationship between theheight of the peak at about 0.7Hz and feelings of quality and sophistication for those
surfaces close together and in the mid-range of component 2, because it is unlikely
that the variance of the responses will allow discrimination of the positions of the
stimuli in semantic space (although the authors have yet to perform this analysis).
The importance of the height of the 0.7Hz peak remains an issue for further testing.
Almost all the surfaces tested demonstrate a peak at between 0.6Hz and 0.8Hz and it
remains to be determined how this arises from the interaction of the finger and the
material from which the surfaces are made.
Component 3 is particularly interesting because of its association with feelings of
warmth. All the tiles were made of the same material and so the effect of thermal
conductivity, even when taking into account the difference in finger to surface contactarea because of different profiles, is likely to be very small. The response might be
accounted for by the other connotations of the word ‘warm’ associated with comfort
and safety, or psycho-physical perception of warmth may depend to some extent on
the profile of the surface.
Surfaces 6 and 7 evoked high responses against component 3. They are similar by not
having distinctive peaks of high relative amplitude. Surface 5, which evoked low
feelings of warmth, does not share these properties. On the other hand, surface 2 does
7/27/2019 An Affective Engineering Study of Vibrational Cues and Affect
http://slidepdf.com/reader/full/an-affective-engineering-study-of-vibrational-cues-and-affect 8/8
not share these properties either, but is close to surfaces 6 and 7 in semantic space
against component 3. Nevertheless, the hypothesis that the presence of distinctive
peaks does not elicit feelings of warmth can be tested in further research.
Other properties of the surface profiles, such as roughness and autocorrelation, remain
to be measured.
CONCLUSIONS
An affective engineering study is reported in which relationships between the
frequency spectra of tactile vibration and affect were examined in the context of car
interiors. A semantic space was generated for surface textures on otherwise identical
black plastic tiles. Vibration measurements were taken when a finger was moved
across each of the surfaces, from which frequency spectra were calculated. The
frequency spectra were compared with the tiles’ positions in semantic space.
There were no clear relationships between the frequency spectra and the tiles’
positions in semantic space. The comparisons suggest the following research
questions which could be tested in future work.
1. Whether there is a relationship between relative amplitude of a frequency peak at between 0.6Hz and 0.8Hz and feelings of quality and sophistication.
2. Whether surfaces with few distinctive frequency peaks of high relative amplitude
feel warmer than surfaces with more distinctive, higher amplitude peaks.
ACKNOWLEDGEMENTS
This work was funded by the UK’s EPSRC (EP/D060079/1) and the EC
(NEST 043157). The views expressed here are the authors’ and not those of the EC.
REFERENCES
Barnes, C.J., Delin, J., Lillford, S. P., Sharoff, S. (2007). Linguistic Support for
Concept Selection Decisions. Artificial Intelligence for Engineering Design, Analysis
and Manufacturing, 21(2), pp.123-135.
Bensmaia, S. J. and Hollins, M. (2003). The vibrations of texture. Somatosensory &
Motor Research, 20(1), pp.33-43.
Kreuger, L. E. (1989). The world of touch by David Katz. Edited and translated by L.
E. Kreuger. Hillsdale: Lawrence Erlbaum Associates, Publishers.
Standex Engraving Group. http://www.standexengraving.com/ Accessed 4 April
2007.
Taylor, M. M., Lederman S. J. (1975). Tactile roughness of grooved surfaces: a model
and the effect of friction. Perception and psychophysics, 64, pp489-494.
Thomas, D. M. H., McEwan J. A. (1988). An application of the repertory grid
method to investigate consumer perceptions of foods. Appetite, 10, pp181-193.